Forum Discussion

Ulf_Zimmermann_'s avatar
Ulf_Zimmermann_
Icon for Nimbostratus rankNimbostratus
May 23, 2012

Renumbering networks and F5 Load Balancers

So our company got aquired a while back. And as life goes all our RFC 1918 IP space use conflicts with our new Overlords. So we have to renumber everything. Our production F5 currently have 783 entries for nodes, virtual servers, rules and selfIPs which need to change.

 

 

 

A mass update would be nice were we could stop both load balancers in the HA setup, manual edit the config files and reboot. But the last time we tried this, it did not work very well due to the HA.

 

 

Business is also asking to reprogram only portions at a time, so a mass update is unfortunatly out.

 

 

 

So now I am looking at suggestions how to do this. I could try and throw something together with perl/expect which uses ssh and tmsh. How about using perl SOAP? Suggestions?

 

 

 

 

 

config
design
  • johns's avatar
    johns
    Icon for Employee rankEmployee
    May 23, 2012
    Are you making any changes to the physical connections or the VLAN information? Not sure what version you have, but you can always make a separate file with just a part of the list and use "b merge" commands to add (instead of removing) it to the bigip.conf on a standby unit, failover to it, then make sure everything works ok, and config sycn. Then you can delete the appropriate section.

     

     

    Not sure why mass-edit did not work, perhaps ARP cache caused a complication?
  • Ulf_Zimmermann_'s avatar
    Ulf_Zimmermann_
    Icon for Nimbostratus rankNimbostratus
    May 23, 2012
    It looked like the mass edit didn't work because HA kicked in and synced the config by itself even I swear it should be doing it.

     

     

    We had changed the on disk files, rebooted the active LB and when it came up, it grabed the config from the standby and all changes were lost.

     

     

    But we have 4 different environments on the production systems (Production, 3x Demo) and Management wants them to be changed seperatly. So I am looking at having to do:

     

     

    Remove Node A from all pools

     

    Delete Node A with old IP

     

    Add Node A with new IP

     

    Add Node A to approriate pools

     

    Repeat for all other nodes

     

    Change IP for all Virtual servers.

     

     

    In total about 783 IP entries to change.

     

  • johns's avatar
    johns
    Icon for Employee rankEmployee
    May 23, 2012
    Unless it is a GTM, it should not be syncing automatically. If it reboots, then it loads from the bigip.conf. If you want to make sure, then just take one of the units offline (put it in maintenance mode, powered off, off the network, etc) and make the mass change and load it manually during maintenance window.
  • Ulf_Zimmermann_'s avatar
    Ulf_Zimmermann_
    Icon for Nimbostratus rankNimbostratus
    May 23, 2012
    Again at this point it doesn't matter as I can't do the mass update, I will have to do the changes one environment at a time without reboot.

     

  • Nathan_Houck_65's avatar
    Nathan_Houck_65
    Icon for Nimbostratus rankNimbostratus
    May 24, 2012
    What version are you running? and are you using Hardware Failover or Network Failover?

     

    Is no downtime a requirement?

     

     

    What about this?

     

    1. On standby unit, backup the bigip.conf and bigip_base.conf files

     

    2. Use a combination of bigpipe/TMSH/sed to edit bigip.conf and bigip_base.conf files as necessary either with manual commands or via sh script

     

    3. do b verify load and then b load

     

    4. Make unit active and do checkouts